Skip to main content

Integrate with AI coding agents

k-ID publishes an official bundle of Agent Skills: small, composable instruction packs that teach an AI coding agent how to integrate k-ID correctly across every jurisdictional regime k-ID supports. They cover COPPA (US), GDPR-Kids (EU), UK AADC, the UK Online Safety Act, Brazil ECA Digital, Australia Online Safety / social media minimum age, and other regional requirements. They work with 35+ AI tools, including Claude Code, Cursor, OpenAI Codex, GitHub Copilot, Gemini CLI, and any other agent that follows the open Agent Skills specification.

The skills live at github.com/kidentify/skills.

Two integration shapes​

The skills support both k-ID integration shapes. The router detects which shape applies from the user's request and loads only the skills that shape needs.

  • Shape A: full CDK integration. Age gate, then session, then consent or verification or threshold, then permissions. For games, social platforms, and multi-feature apps with persistent per-user state and multiple gated features. Uses 5–7 of the skills.
  • Shape B: standalone AgeKit+. A single call to /age-verification/perform-access-age-verification, an iframe for the user to complete verification, and a webhook or polled result. No age gate, no session, no permissions. For 18+ sites (UK OSA), age-restricted downloads, Australia social-media minimum-age checks, or any single age-proof decision. Uses 2–3 of the skills.

Two UI approaches in custom integrations​

Within Shape A, k-ID supports two ways of rendering the age gate and parental-consent flow, and the skills cover both. See Choose integration for the canonical comparison. The short version:

  • Custom UX workflows (the default). Build your own age-gate and consent UI and call /age-gate/check, /challenge/send-email, and the rest directly. Produces the best-looking, most brand-integrated experience and works on every platform (web, Unity WebGL, consoles, native desktop). Recommended for production integrations.
  • Widgets (fast-path fallback). Pre-built k-ID iframes for the age gate, the end-to-end flow (age gate + parental consent + data notices + permissions + parental preferences), manage-permissions, and data notices. They handle jurisdiction-appropriate age collection and initiate parental-consent challenges automatically. Use when the integration must be small, simple, and fast to ship.

The k-id-age-gate and k-id-consent-and-challenges skills both document the custom path (Pattern A) and the widget path (Pattern B). Just tell the agent what you want, and it picks the right one.

What the skills give you​

Eight focused skills that compose:

SkillWhat it teaches the agentShape
k-id-integrationRouter that detects shape and picks the right sibling skills.Both
k-id-age-gateBuilds the age gate with either a fully custom slider calling /age-gate/check (Pattern A: the default, best-looking and most brand-integrated) or the k-ID age-gate widget (Pattern B: fast-path fallback for small, simple integrations). Entry point to every Shape A flow.A
k-id-consent-and-challengesGUARDIAN-managed parental consent for minors. Use a custom consent screen with QR + OTP + email + direct link and top-level polling outside the modal (Pattern A: the default) or the end-to-end / manage-permissions widget (Pattern B: fast-path fallback).A
k-id-age-verificationAge verification and assurance: standalone AgeKit+ (Pattern 1), unverified-adult in a session (Pattern 2), and per-permission verifiedAgeThreshold flows (Pattern 3, UK OSA 18+, Brazil ECA Digital, Australia social media minimum age).Both
k-id-sessions-and-permissionsSession handling, permission-gated UI controls, /session/upgrade shape, and verifiedAgeThreshold handling.A
k-id-webhooksHMAC-SHA256 signature verify against the raw body, idempotency, event handlers.Both
k-id-server-trust-boundaryAPI key placement, server proxy, pre-flight checks.Both
k-id-mobile-nativeiOS / Android / Unity platform age signals and in-app browsing.A (usually)

When the agent activates k-id-integration, it reads the user's task and pulls in the right combination of feature and cross-cutting skills, so you don't install them individually.

Before you install​

  • You need a k-ID product in Compliance Studio with an API key and webhook secret. See the API authentication guide.
  • The skills reference API shapes from this documentation site. If you've blocked outbound network access for your AI agent, allow docs.k-id.com.

Install by tool​

All installation paths below get the same eight skills into your agent's context. Pick the row that matches your tool.

Claude Code (fastest)​

Inside Claude Code, run:

/plugin marketplace add github.com/kidentify/skills
/plugin install k-id-skills@kidentify

The plugin registers all eight skills at once. Confirm with /plugin list. The marketplace URL must be added as a git-style reference (github.com/... or https://github.com/...); if you see a schema error on plugins[].source, update the skills repository .claude-plugin/marketplace.json so each plugin source is a relative path starting with ./ (see Claude Code plugin marketplaces).

Cursor​

Clone the skills/ directory from the repository into your project's .agents/skills/ directory (or use .cursor/skills/). Cursor discovers project skills from those paths automatically; for user-wide skills, use ~/.agents/skills/ or ~/.cursor/skills/. See Cursor Agent Skills.

mkdir -p .agents/skills
git clone --depth 1 https://github.com/kidentify/skills.git /tmp/k-id-skills
cp -R /tmp/k-id-skills/skills/* .agents/skills/
rm -rf /tmp/k-id-skills

Reload Cursor and open a chat to confirm the skills appear in the agent's available tools. Cursor picks up new skills without a restart in most cases.

OpenAI Codex / Codex CLI​

Codex discovers skills under .agents/skills/ from your working directory up to the repository root, and user-wide skills under ~/.agents/skills/. Those are the paths OpenAI documents for the Codex CLI and IDE extension; ~/.codex/skills/ can still be read for backward compatibility but isn't the primary location. For per-project setup, use the same git clone + cp recipe as Cursor. For global:

mkdir -p ~/.agents/skills
git clone --depth 1 https://github.com/kidentify/skills.git /tmp/k-id-skills
cp -R /tmp/k-id-skills/skills/* ~/.agents/skills/
rm -rf /tmp/k-id-skills

See Codex Agent Skills.

GitHub Copilot / Visual Studio Code​

GitHub Copilot in Visual Studio Code loads project skills from .github/skills/, .claude/skills/, or .agents/skills/ (see Use Agent Skills in Visual Studio Code). Personal skills can live under ~/.copilot/skills/, ~/.claude/skills/, or ~/.agents/skills/. Copy the repository skills/ tree into the project path you prefer; .github/skills/ matches the GitHub-documented layout for repo-scoped skills:

mkdir -p .github/skills
git clone --depth 1 https://github.com/kidentify/skills.git /tmp/k-id-skills
cp -R /tmp/k-id-skills/skills/* .github/skills/
rm -rf /tmp/k-id-skills

Use /skills in Chat to confirm they appear. Optional chat.agentSkillsLocations can add more roots if your team uses a non-default folder.

Gemini CLI​

Gemini CLI discovers user skills in ~/.gemini/skills/ or ~/.agents/skills/ (the latter takes precedence if both exist at the same tier), and workspace skills in .gemini/skills/ or .agents/skills/. See Gemini CLI Agent Skills. Global install:

mkdir -p ~/.gemini/skills
git clone --depth 1 https://github.com/kidentify/skills.git /tmp/k-id-skills
cp -R /tmp/k-id-skills/skills/* ~/.gemini/skills/
rm -rf /tmp/k-id-skills

For a repo-local install shared with the team, use .agents/skills/ or .gemini/skills/ the same way as the Cursor recipe. You can also run gemini skills install https://github.com/kidentify/skills.git (see Gemini CLI skills management).

Other tools and air-gapped environments​

Any AI agent that follows the Agent Skills specification can consume these skills. The Agent Skills client showcase lists 35+ compatible tools. Each tool documents where it reads skills from. Once you know the directory, the copy recipe is the same.

If outbound internet is blocked, download the ZIP via the "Code" button on GitHub, transfer it to your environment, and extract the skills/ directory into the tool skills path for your agent.

Last verified

These install paths were checked against Cursor, OpenAI Codex, Visual Studio Code with GitHub Copilot, and Gemini CLI documentation on 2026-05-15. If a tool changes its skills directory, follow that product documentation and the Agent Skills client showcase.

How to use the skills​

Once installed, just describe your task to the agent in natural language:

  • "Integrate k-ID into a Next.js app."
  • "Add parental consent with QR, OTP, and email."
  • "Verify the webhook handler against the Compliance Studio Test event."
  • "On iOS, read the declared age range before the age gate."

The router skill (k-id-integration) activates first, reads the request, and hands off to the right specialized skill. You don't need to remember skill names.

For deeper questions about the patterns the skills encode, read the relevant concept page:

Design principles behind these skills​

  • Doc-first. API request and response shapes come from this documentation site. Skills encode integration patterns and known pitfalls (not endpoint reference data), so a shape change on the k-ID side never leaves a skill out of sync.
  • Invariants inline. Each SKILL.md includes a "Gotchas" section with the rules that prevent real, shipped bugs, each paired with a one-line "why."
  • Calibrated prescriptiveness. Fragile operations (API body shapes, signature verification) are specified exactly. UI adaptation and framework choice are left to you.
  • Validate at the end. Every feature skill ends with a verification checklist so the agent confirms the integration is actually working before declaring done.